Daily
Thursday, 17 September 2026
Top 50 · open 01–03
01Nvidia announces native GPU programming in Rustmust NVIDIA is compiling Rust kernels to PTX instead of wrapping someone else's C++. Two tracks: cuda-oxide is a custom rustc backend through Pliron and LLVM (pinned nightly, early alpha, DisjointSlice so classic aliasing fails at compile time), and cutile-rs is Tile IR on stable 1.89, already on crates.io and in Hugging Face Grout and mistral.rs. Neither is production-ready; I opened it because kernel-in-Rust with actual safety types is the thing the CUDA C++ people said would never ship.
claiirfunnyHN
The launch is checked rather than trusted. Damn even Nvidia is putting out fully Claude-written articles.
02Training a 4B model to produce 81% faster query plans than Postgresmust Rohan Bansal took a 4B Qwen distill that couldn't even speak the planner harness, distilled Astra trajectories into it, then ran agentic GRPO against live Postgres — 2× H100 for vLLM, four containers on a desk for measurement. Best-of-15 on the 113 JOB queries is 1.81× geometric mean versus the default planner and 44.7% less summed latency, about $1,200 all-in. I opened it for the measurement rig: Linux page cache was lying to the reward, so he had to denoise Postgres before the model could learn.
malisperfunnyHN
Funnily enough, you could replace "LLM query planner" with just "query planner" and this comment would still hold true
03Breaking the 1.58-bit Barrier for Ternary LLMs Georganas, Heinecke, and Dubey measured 29 ternary LLMs and found zeros eat up to 51.5% of the weights, so log₂(3)≈1.58 was always a bit of a fiction. BITCOS is a presence bitmap plus compacted signs at 2−z bits, beating five-trit packing on 26 of 29 models down to 1.485 bits/weight, with AVX-512 / AVX2 / Xe2 unpackers and up to 1.28× kernel and 1.27× GPU decode. Intel packing paper, not a new architecture — I glanced because the 1.58-bit slogan finally met an entropy measurement.
04Xiaomi Mimo 2.6 live post-training dashboard 05Backups Aren't Simple 06Small programming tricks 07The engineering behind the US Strategic Petroleum Reserve 08OpenSpec – A lightweight and configurable AI spec framework 09Reversing Factorio's RNGmust 10Performance Improvements in .NET 11 11Australia says it could follow Canada in forging deeper ties with EU 12AWS says it can't restore some data from mideast facilities struck by Iran 13Japan's book scene is moving from bookstores to libraries 14Reverse-engineered Jev-like model 15How good are frontier models at physics?must 16Show HN: An e-ink frame that hears birds and draws them as 1800s illustrations 17Accurate Models of AMD Matrix Coresmust 18Mistral X Mozilla: Private, Multilingual AI Browsing 19Anecdotally, programmers dislike "reduce" 20Why Does the Universe Expand? 21Dream-RSI: Recursive Self-Improvement through Evolving Worlds 22Anatomy of a Texture 23HarnessTax: How Much Does the Harness Matter for Coding Agents? 24Training Text-to-Image Models 3.6× Faster 25William Buckland's Theology of Geology 26Destroy After Reading: photocopiers,cheap paper and DIY gave metal it's look 27Hackers Got Inside a Flock Camera 28The DeepMind Institute 29WalShadow: Sub-second Postgres replication to ClickHouse from physical WAL 30Kyber (YC W23) Is Hiring a Forward Deployed Engineer 31Vectorized and performance-portable Quicksort (2022) 32I replaced my brown-noise browser tab with a menu bar app 33Tell the speakers that you liked their talks 34Autism is genetic – and why this information matters now 35The Siberian Ice Maiden and the Scythian World 36Show HN: AttaLambda: a language where types and data are made of untyped lambdas 37Common Crawl Data Stored on a Hugging Face Bucket 38A warning about 'model welfare' 39Show HN: Restarted – a 2026 remake of the classic 2015 startup generator 40Douglas Adams and the exterminated Doctor Who adventure 41How big are factorials? 42The Google Play app review process now regularly takes longer than a week 43Claude Cowork and chat are now one Claude 44My website charged AI agents a penny per page. I watched Claude pay it 45Introducing System One Models and Jev 46Paramount 'Leaks' Study Saying California Will Suffer If Its Merger Is Blocked 47Show HN: How Stale Is Your AI? Release age and training cutoff for 20 models 48Salesforce Global Outage 49Write Linters and Tools Before Code 50Doing Everyone Else's Job